Re: Index compression vs. table compression

Thomas Kyte wrote:
> In article <csv1if$fnr$1_at_news-02.connect.com.au>, Howard J. Rogers says...
>
>>Richard Foote wrote:
>>
>
> ...
>
>
>>But, Richard, in the sentence which Tom has been elaborating on, and
>>which was made in response to Jonathan, I made no mention of lookup tables.
>>
>
>
> we are exceedingly close to beating a rather dead horse, but -- as for putting
> words into mouths of others -- my posting had verbaitim cut and pastes which
> I'll tag with initials here:
>
>
> Author: "Howard J. Rogers" <hjr_at_dizwell.com>
> Date: Fri, 31 Dec 2004 10:40:48 +1100
> HJR>>>is then a further problem: how is the optimiser likely to read small,
> HJR>>>useful, lookup tables?.. er, via a FTS, probably, if they are genuinely
> HJR>>>small.

That was part of a discussion with Rick.

>
> Author: "Jonathan Lewis" <jonathan_at_jlcomp.demon.co.uk>
> Date: Fri, 31 Dec 2004 08:14:28 +0000 (UTC)
> JPL>> Not if they're being used for doing lookups, I hope.
> JPL>>
That is Jonathan's intervention

>
> Author: "Howard J. Rogers" <hjr_at_dizwell.com>
> Date: Sat, 01 Jan 2005 03:11:59 +1100
> HJR> Why?
> HJR> A small table is always likely to be read via a FTS using CBO. Even for a
> HJR> single key lookup...

And that is my reply to Jonathan.

> You actually did use "lookup tables".

Not in the part you actually chose to quote in your earlier post, I didn't! It's there in black and white! I did NOT use the noun 'lookup table' when replying to Jonathan (which is the bit you chimed in on).

I mentioned lookup tables when discussing a point with Rick. Jonathan then intervened with a generalised comment about 'lookups', and I replied *to him* with an even more generalised comment about 'single key lookups'.

That the specific noun 'lookup table' appeared earlier on, in a converstation with someone else, does not mean that is what I was talking about when replying to Jonathan.

I can assure you that if I had meant to say lookup table *at that point*, I would have done so.

>You said lookup tables, Jonathan said
> lookup table, you changed it to "single key lookup" after the fact, but we were
> in fact discussion lookup tables -- which will use an index, not a full table
> scan.

No, "we" weren't, because *I* was discussing something *with Jonathan*.

And you quoted *that* part of the conversation, and made the assumption that a single key lookup was the same thing as I'd been talking to Rick about earlier.

I *showed* you what I meant by a single key lookup. And no, I didn't hurriedly knock that up in a desperate attempt to cover my ass. And equally no, I am not that desperate to be thought infallible that I wouldn't simply fess up and admit an error *if that is what I had actually made*.

But I know what I wrote, and I know what I meant; and if you, and others, would just read the words I did use, and not interpolate words that aren't there, then that meaning should be clear enough to others too. If it isn't, then let's clarify. But let's clarify what I actually said, not foist meanings on to it that were never there.

> So, when talking about lookup tables, a keyed read of a single row, indexes
> rule.
>
> When talking about large range scans, the optimizer will choose a full scan or
> range scan depending on many factors including the estimate number of rows to be
> retrieved, clustering factor, optimizer_index_* settings, multi-block read
> counts etc.... It is never an "always" situation, it is always "it depends"
> situation.

If that is directed at me, I didn't say it was an "always" situation, either. "Always likely" is not the same thing as "always".

I don't quite understand the current fashion to say I said things when I didn't. I am reduced to having to point out the difference between what I did actually say, and what I am supposed to have said, which has the unfortunate effect of making me sound like a verbal pedant of the worst order.

Regards
HJR Received on Sun Jan 23 2005 - 12:56:59 CST